aboutsummaryrefslogtreecommitdiffstats
blob: 00ef898646608cb49cec7c263b2b39a19f0c829f (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
/******************************************************************************
 * Copyright (c) 2013-2014, Texas Instruments Incorporated - http://www.ti.com/
 *   All rights reserved.
 *
 *   Redistribution and use in source and binary forms, with or without
 *   modification, are permitted provided that the following conditions are met:
 *       * Redistributions of source code must retain the above copyright
 *         notice, this list of conditions and the following disclaimer.
 *       * Redistributions in binary form must reproduce the above copyright
 *         notice, this list of conditions and the following disclaimer in the
 *         documentation and/or other materials provided with the distribution.
 *       * Neither the name of Texas Instruments Incorporated nor the
 *         names of its contributors may be used to endorse or promote products
 *         derived from this software without specific prior written permission.
 *
 *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 *   AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 *   IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 *   ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
 *   LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 *   CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 *   SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 *   INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 *   CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 *   ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
 *   THE POSSIBILITY OF SUCH DAMAGE.
 *****************************************************************************/
#define __CL_ENABLE_EXCEPTIONS
#include <CL/cl.hpp>
#include <iostream>
#include <fstream>
#include <cstdlib>
#include "ocl_util.h"
#include "conv.dsp_h"

//#define VERBOSE
using namespace cl;
using namespace std;

/*----------------------------------------------------------------------------------------------------------------------*/
static int oclconv_imgproc(char *kernelName, unsigned char *data_in, unsigned char *data_out, int width, int height, int sstride, int dstride)
{
   cl_int err     = CL_SUCCESS;
   int  bufsize   = sstride * height * sizeof(unsigned char);
#ifdef VERBOSE
   ofstream logfile;
   logfile.open ("/home/root/oclconv_log.txt", ios::out | ios::app); 
   logfile << "Entered oclconv_test, width=" << width << " height=" << height << " dstride=" << dstride << " sstride=" << sstride << '\n';
   logfile.close();
#endif

   try 
   {
     Context context(CL_DEVICE_TYPE_ACCELERATOR);
     std::vector<Device> devices = context.getInfo<CL_CONTEXT_DEVICES>();
     devices.resize(1); // resize to 1 since we are only running on 1 DSP

     Buffer bufA   (context, CL_MEM_READ_ONLY,  bufsize);
     Buffer bufDst (context, CL_MEM_WRITE_ONLY, bufsize);

     Program::Binaries binary(1, make_pair(conv_dsp_bin,sizeof(conv_dsp_bin)));
     Program           program = Program(context, devices, binary);
     program.build(devices);
     Kernel kernel(program, kernelName);
     kernel.setArg(0, bufA);
     kernel.setArg(1, bufDst);
     kernel.setArg(2, width);
     kernel.setArg(3, height);
     kernel.setArg(4, dstride);
     kernel.setArg(5, sstride);

     Event ev1,ev2,ev3,ev4;

     CommandQueue Q(context, devices[0], CL_QUEUE_PROFILING_ENABLE);

     Q.enqueueWriteBuffer(bufA, CL_FALSE, 0, bufsize, data_in, NULL, &ev1);
     Q.enqueueTask (kernel, NULL, &ev3);
     Q.enqueueReadBuffer (bufDst, CL_TRUE, 0, bufsize, data_out, NULL, &ev4);
   }
   catch (Error err) 
   { cerr << "ERROR: " << err.what() << "(" << err.err() << ")" << endl; return -1; }
#ifdef VERBOSE
   logfile.open ("/home/root/oclconv_log.txt", ios::out | ios::app); 
   logfile << "Success!" << endl; 
   logfile.close();
#endif
   return 0;
}
/*----------------------------------------------------------------------------------------------------------------------*/
static bool          canny_first_call = true;
static Context       canny_ctx(CL_DEVICE_TYPE_ACCELERATOR);
static CommandQueue *canny_Q;
static Buffer       *canny_gradX, *canny_gradY, *canny_mag, *canny_scratch, *canny_numItems;
static Kernel       *canny_K;
static Buffer       *canny_input, *canny_output;

/******************************************************************************
 * Canny Edge Detection - called on ARM, but algorithm dispatched to 1 DSP
 *
 * Note: Assumes arguments are invariant from call 1 to call N. If this is 
 *   not the case, then move buffer creation back to the every frame section
 *   rather than being cached in frame 0.
 *
 * Note: Also assumes total size is not overly large as it allocates temp 
 *       buffers in MSMC
 *****************************************************************************/
static int ocl_canny(unsigned char *data_in, unsigned char *data_out, unsigned short height, unsigned short width)
{
    int numelem = (int)height*(int)width;
    try 
    {
        Event canny_ev, canny_ev1, canny_ev2;
        /*---------------------------------------------------------------------
        * Cache as much OpenCL plumbing on the first call, so the cost is not 
        * repeatedfor every frame.
        *--------------------------------------------------------------------*/
        if (canny_first_call)
        {
            canny_first_call = false;

            std::vector<Device> devices = canny_ctx.getInfo<CL_CONTEXT_DEVICES>();
            devices.resize(1); // resize to 1 since we are only running on 1 DSP
            canny_Q = new CommandQueue(canny_ctx, devices[0]);

            canny_input   = new Buffer(canny_ctx, CL_MEM_READ_ONLY, numelem);
            canny_output  = new Buffer(canny_ctx, CL_MEM_WRITE_ONLY, numelem);
            canny_gradX   = new Buffer(canny_ctx, CL_MEM_WRITE_ONLY, numelem*sizeof(short));
            canny_gradY   = new Buffer(canny_ctx, CL_MEM_WRITE_ONLY, numelem*sizeof(short));
            canny_mag     = new Buffer(canny_ctx, CL_MEM_WRITE_ONLY, numelem*sizeof(short));
            canny_scratch = new Buffer(canny_ctx, CL_MEM_WRITE_ONLY, numelem);
            canny_numItems= new Buffer(canny_ctx, CL_MEM_WRITE_ONLY, sizeof(int));

            /*---------------------------------------------------------------------
            * Compile the Kernel Source for the devices
            *--------------------------------------------------------------------*/
            Program::Binaries binary(1, make_pair(conv_dsp_bin, sizeof(conv_dsp_bin)));
            Program program(canny_ctx, devices, binary);
            program.build(devices);
            canny_K = new Kernel(program, "canny_tiocl");

            canny_K->setArg(0, *canny_input);
            canny_K->setArg(1, *canny_gradX);
            canny_K->setArg(2, *canny_gradY);
            canny_K->setArg(3, *canny_mag);
            canny_K->setArg(4, *canny_output);
            canny_K->setArg(5, *canny_scratch);
            canny_K->setArg(6, *canny_numItems);
            canny_K->setArg(7, width);
            canny_K->setArg(8, height);
        }

       canny_Q->enqueueWriteBuffer(*canny_input, CL_FALSE, 0, numelem, data_in, NULL, &canny_ev1);
       canny_Q->enqueueTask(*canny_K, 0, &canny_ev);
       canny_Q->enqueueReadBuffer (*canny_output, CL_TRUE, 0, numelem, data_out, NULL, &canny_ev2);
    }
    catch (cl::Error err)
    {
       cerr << "ERROR: " << err.what() << "(" << err.err() << ")" << endl;
       return (-1);
    }
    return 0;
}
/*----------------------------------------------------------------------------------------------------------------------*/

#ifdef __cplusplus
extern "C" {
#endif
int oclconv_kernel(int kernel_type, int filter_size, 
                   unsigned char *data_in, unsigned char *data_out, 
                   int width, int height, int dstride, int sstride)
{
  int retval = -1;
  switch(kernel_type)
  {
    case 0: /* Median */
      if(filter_size == 5) { 
        retval = oclconv_imgproc("Median2x2", data_in, data_out, width, height, sstride, dstride);
      } else if(filter_size == 9) { 
        retval = oclconv_imgproc("Median3x3", data_in, data_out, width, height, sstride, dstride);
      }
      break;
    case 1: /* Sobel */
      if(filter_size == 9) { 
        retval = oclconv_imgproc("Sobel3x3", data_in, data_out, width, height, sstride, dstride);
      }
      break;
    case 2: /* conv */
      if(filter_size == 25) { 
        retval = oclconv_imgproc("Conv5x5", data_in, data_out, width, height, sstride, dstride);
        return 0;
      }
      break;
    case 3: /* vlib canny */
      /* filter size is ignored */
      retval = ocl_canny (data_in, data_out, width, height); /* input and output stride assumed to be == width */
      break;
    default:
      break;
  }
  return retval;
}
#ifdef __cplusplus
}
#endif